Turn Your AI Assistant Into a Pulumi Expert

Finally, you can stop switching between Claude and your terminal every time you need to preview a stack change or check deployment outputs. The Pulumi MCP Server bridges that gap, giving your AI assistant direct access to Pulumi operations so you can manage infrastructure through natural conversation.

What This Solves

If you're already using Pulumi, you know the drill: write some infrastructure code, run pulumi preview to check changes, maybe adjust something, run pulumi up, check outputs, rinse and repeat. Now imagine doing all of that without leaving your AI chat.

Instead of:

# Terminal 1
pulumi preview --stack dev
# Copy output, paste into Claude
# Claude suggests changes
# Back to terminal
pulumi up --stack dev
# Copy outputs, back to Claude for next steps

You get:

You: "Preview my changes on the dev stack, then if it looks good, deploy it"
Claude: *runs pulumi preview, analyzes output, runs pulumi up, shows results*

Real Workflow Improvements

Infrastructure Troubleshooting: Ask Claude to check stack outputs, compare with expected values, and suggest fixes - all while it has live access to your actual stack state.

Resource Discovery: "What inputs does the AWS RDS cluster resource accept?" Claude can query the Pulumi registry directly instead of you hunting through docs.

Smart Deployments: The deploy-to-aws command analyzes your application code and generates appropriate Pulumi infrastructure. Point it at a Node.js app, and it'll create Lambda functions, API Gateway, and necessary IAM roles.

Stack Management: Compare outputs across environments, validate configurations, and even perform bulk operations across multiple stacks through natural language commands.

Key Capabilities

• Live Pulumi Operations: preview, up, stack-output commands run directly through your AI assistant
• Registry Integration: Query any Pulumi provider for resource documentation and schemas
• Intelligent Deployment: Automatic infrastructure generation based on application analysis
• Multi-Transport Support: Works via stdio for Claude Desktop or HTTP/SSE for web applications
• Docker Ready: Run containerized to avoid local Node.js setup

Quick Integration

Claude Desktop Setup

Add to your MCP configuration:

{
  "mcpServers": {
    "pulumi": {
      "command": "npx",
      "args": ["@pulumi/mcp-server@latest", "stdio"]
    }
  }
}

Claude Code Setup

claude mcp add -s user pulumi -- npx @pulumi/mcp-server@latest stdio

Docker Alternative

{
  "mcpServers": {
    "pulumi": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "-v", "~/projects:/app/projects", "pulumi/mcp-server:latest", "stdio"]
    }
  }
}

Production Use Cases

Code Reviews: Have Claude preview infrastructure changes and explain the impact before merging PRs.

Deployment Automation: Create workflows where Claude analyzes application changes, suggests infrastructure updates, previews them, and deploys after approval.

Multi-Environment Management: "Deploy this to staging, then compare the outputs with production to make sure they match."

Resource Optimization: "Check all my AWS resources and suggest cost optimizations" - Claude can analyze your actual deployed infrastructure.

The server runs locally and uses your existing Pulumi CLI installation, so you maintain full control over credentials and access. Your infrastructure operations just become as conversational as asking Claude to write code.

This isn't about replacing Pulumi's CLI - it's about making your existing Pulumi workflow more intelligent and reducing the friction between thinking about infrastructure and actually managing it.